doc/articles/explore-your-app-ai-agents.md
Now that you've created your Uno Platform app, you're setup to have your agent help you develop features. In the agent chat, ask the following:
[!IMPORTANT] The MCP tools are best used by agents when using premium models like Claude 4.5, GPT-5-Codex or GPT-5.
Add a button and a textblock in my main page. Have a button click
display the current date in the new TextBlock. Make sure to validate
the app's runtime behavior using your tools.
Your AI agent will act and use the tools provided by Uno Platform.
Uno Platform provides two MCP (Model Context Protocol) servers, the Remote Uno MCP and the Local App MCP.
For more detailed information, you can read further on the tools offered by Uno Platform MCPs.
This is hosted publicly and provides:
The Uno App MCP is a runtime service for the Uno Platform that allows large language models (LLMs) or other agent-based tools to intelligently interact with live Uno applications across platforms such as Windows, WebAssembly, macOS, iOS, Android, and Linux, in a real runtime context.
Rather than relying on static UI captures or screenshots, Uno App MCP exposes structured application state (like the Visual Tree, data context, and control properties), enabling rich, meaningful automation and inspection.
It allows tools and agents to:
In other words: while traditional Uno tests are written and maintained by developers, Uno App MCP opens the door for AI agents or other automation systems to drive and reason about the app autonomously, with deeper context and less manual scripting, across all Uno-supported platforms.
In your agent, there are some phrases that can be used to nudge the agent to use the tools.
[!NOTE] Your agent will execute tools from our MCPs, you will need to approve them at your convenience. [!NOTE] In Visual Studio 2022/2026 MCPs might not be enabled automatically. Make sure to click the "tools" icon in the chat window to check both Uno Platform MCPs.
If the Uno App MCP is missing, see how to make the App MCP appear in Visual Studio.
To ask the agent to explain what MVUX is:
Can you use the uno platform documentation to explain what MVUX is?
To ask the agent to take a screenshot and analyze it:
Can you take a screenshot of the running app and describe what you see?
[!NOTE] Some models don't yet support analyzing images, particularly in Visual Studio. Saving a screenshot on disk is always available.
To ask the agent to click on a specific element on screen:
Can you click the second button on the left, close to the middle of the screen?
To ask the agent to click on a specific element on screen:
Can you type "Hello Uno Platform, the MCP is talking!" in the third textbox on the screen?
To ask to delete all the text of a textbox
Can you select all the text of the second textbox and delete all of it?
To analyze the DataContext of the app:
Can you tell me if there's an object of type "XYZ" in the app's
DataContext, starting from the grid in the middle of the screen?
To create a more specialized Uno Platform agent:
/mcp.uno.init
[!NOTE] The default command is
/mcp.uno.init, but depending on the agent you are using, it might be named differently (such as/uno.init).
Which primes the agent with Uno Platform's best practices.
Then ask the agent to build a new feature:
Let's add a new page in my app that shows the "About" information.
All those phrases will hint the model to use one or more of the tools to answer your request.
You can find additional information about troubleshooting AI Agents in our docs.